Skip to contents

Design effect from clustering

Source: R/subpop_design.R
morie_deff_cluster.Rd

The standard cluster design effect \(deff = 1 + (m - 1)\rho\), where \(m\) is the average number of completed interviews per cluster and \(\rho\) the intracluster correlation. When a survey is clustered by community, the effective sample size is the nominal size divided by this quantity, and planning that ignores it understates the sample required.

Usage

morie_deff_cluster(m, icc)

Arguments

m

Average number of completed units per cluster. Must be at least 1.

icc

Intracluster correlation \(\rho\). May be negative, but must not produce a negative design effect.

Value

A single numeric design effect.

Details

Use morie_design_effect instead when unequal weights, not clustering, are the source of the loss; the two sources multiply.

See also

morie_neff_cluster, morie_design_effect

Examples

# 25 interviews per community, ICC 0.02
morie_deff_cluster(25, 0.02)
#> [1] 1.48